XPath Cheatsheet - Axes, Predicates & Functions

They differ in semantics. / is an absolute path from the document root that follows one definite branch downward, while // is shorthand for "search descendants recursively from anywhere" and each evaluation walks the remaining subtree. So /bookstore/book inspects the book elements under a single bookstore, whereas //book has to examine every node in the document. On small documents the difference is negligible; once nodes reach into the tens of thousands, a // nested inside several levels multiplies the traversal cost. Prefer an exact path when the depth is known, fall back to // only when it is not, and place it once at the start of an expression with relative paths expanding from there.

Indexing starts at 1, so //li[2] is the second matching li rather than the third — the opposite of array-index intuition. Two details cause most mistakes. First, the index counts within the filtered result set, so predicates come before the index: //li[@class="x"][2] is the second li carrying that class, and reversing the order changes the meaning entirely. Second, a bracket may hold several conditions meaning all must hold, and comparisons like @price>35 require the attribute to be convertible to a number — otherwise the expression is invalid rather than quietly returning an empty set.

Real HTML classes are almost always composite: class="btn btn-primary active" holds three classes at once. contains(@class, "active") tests whether the attribute value includes a substring, so it hits that form, whereas exact equality only matches when the whole attribute value is precisely active, which in practice almost never happens. contains is therefore the default choice for scraping. The cost is that it also matches active-tab and inactive by pure containment, so it is looser than exact matching without being precise; combine it with another condition when you need strictness.

1.0 has no regular-expression function — that is a version boundary, not a syntax mistake. There are three workable substitutes: contains combined with length checks for containment matching, which covers most real cases; substring for positional trimming, which suits fixed formats such as codes and dates; or switching to a processor that implements XPath 2.0, where regex-style matching functions exist. Confirm the runtime version before choosing, or the same expression behaves differently in two environments.

Bind the root node to a variable once, then take relative paths beneath it. Once you hold the product-list node, every field expression is evaluated within that small subtree instead of each field restarting a // traversal across the entire tree. For configuration files there is a related habit: prefer element names over attribute values in paths, because element names are usually the public contract of the schema while attributes are more likely implementation detail — and attributes are what change first.